# Moltis

> A personal AI assistant built in Rust with local LLM support, multi-channel access, and sandboxed execution capabilities.

Moltis is a self-hosted personal AI assistant built in Rust that runs as a single binary with no runtime dependencies. It supports local LLMs with automatic download and setup, multiple cloud providers, and offers multi-channel access through Web UI, Telegram bot, and JSON-RPC API. The assistant features sandboxed execution environments for safer automation and long-term memory with hybrid vector and full-text search.

- **Single Binary Deployment** - Download and run one self-contained binary with no runtime dependencies, available via Homebrew, Cargo, Docker, or direct download for Debian, Fedora, Arch Linux, Snap, and AppImage.

- **Local LLM Support** - Run your own models locally with automatic download and setup included, supporting offline operation with GGUF embeddings and provider fallback chains.

- **Multi-Channel Access** - Connect through Web UI, Telegram bot, JSON-RPC API, or Mobile PWA with push notifications and multi-device sync capabilities.

- **Sandboxed Execution** - Run browser sessions and shell commands in isolated Docker containers or Apple Container isolation for safer automation with SSRF protection.

- **Long-term Memory** - Hybrid vector and full-text search enables your agent to remember context, with file watching, live sync, and session export features.

- **MCP Server Support** - Extend functionality with MCP tool servers, skills, and hooks supporting Stdio or HTTP/SSE with auto-restart capabilities.

- **Voice Integration** - Talk to your assistant with multiple cloud and local TTS/STT providers for speech-to-text and text-to-speech functionality.

- **Security Features** - Includes passkeys (WebAuthn), scoped API keys, secrets zeroed on drop, human-in-the-loop approval, origin validation, and no unsafe code.

- **Observability** - Built-in Prometheus metrics, OpenTelemetry tracing, structured logging, per-provider charts, and real-time WebSocket monitoring.

To get started, install Moltis using the quickstart script, Homebrew, Cargo, or Docker. Open the web interface at localhost:13131 to complete setup with passkey or password authentication. Configure your LLM providers and channels, then start chatting with voice, map, and tool use capabilities.

## Features
- Single binary deployment
- Local LLM support with automatic download
- Multi-channel access (Web UI, Telegram, API)
- Sandboxed browser execution in Docker
- SSRF protection
- Streaming-first responses
- MCP server support
- Skills and hooks system
- Long-term memory with hybrid search
- Pi-inspired self-extension
- Voice support (TTS/STT)
- Passkeys (WebAuthn) authentication
- Scoped API keys
- Human-in-the-loop approval
- Cron job scheduling
- Prometheus metrics
- OpenTelemetry tracing
- Session export
- Push notifications
- Multi-device sync

## Integrations
OpenAI, GitHub Copilot, OpenAI Codex, Telegram, Docker, DigitalOcean, Fly.io, Prometheus, OpenTelemetry

## Platforms
WINDOWS, MACOS, LINUX, WEB, API

## Pricing
Open Source

## Links
- Website: https://www.moltis.org
- Documentation: https://docs.moltis.org/
- Repository: https://github.com/moltis-org/moltis
- EveryDev.ai: https://www.everydev.ai/tools/moltis